of a 
ndly, 
object 
antics 
ng of 
tities 
plicit 
patial 
1 in the 
tance of 
is the 
, whose 
se. The 
ypes to 
its that 
^sented 
ication 
trained 
pes of 
riterion 
ity has 
rves as 
‘mation 
that he 
base is 
has to 
on the 
ints are 
s of the 
ferring 
location 
; not be 
ntegrity 
r from 
ye, that 
es. The 
specification of integrity constraints is application 
dependent; it is the aimed use of the database that 
determines the constrained objects and the way they 
are constrained. 
1.2 Previous studies and outline of this study 
The importance of consistency is emphasized in for 
example (NCDCDS, 1987, part III, pp.5-6). Description of 
consistency tests is proposed as a part of the quality 
report of the data. The description of consistency is 
thus a part of metadata describing a geographical 
database. In (White, 1984) questions regarding the 
consistency of a multipurpose geographic data system 
are represented. In (Laurini & Milleret-Raffort, 1991), 
correct geometric construction of spatial objects is 
emphasized, and in (Molenaar, 1991) rules are listed 
which can be used to check the consistency of a 
database. 
In this study, emphasis is on object-oriented modelling 
of geographical entities, that is, definition of 
geographical object types in object-oriented databases, 
and enforcement of their integrity. Geometrical object 
model is presented which can be used in describing the 
geometrical structure of real world entities. The 
structural features of the object model capture the 
geometry of an object, while the procedural features of 
the model enforce the structural integrity of an object. 
Additionally, a technique is proposed by which user 
defined constraints may be specified to a geographical 
database. 
The concepts of object-oriented programming and 
databases are not explained in this paper. The author 
things that the concepts such as class, attribute, 
method, inheritance etc. are generally quite well 
known. An earlier paper by author (Kemppainen & 
Sarjakoski, 1990) is a review of concepts of object- 
oriented programming and databases, and their 
application to the development of geographical 
information systems. 
In the next section (2), the relationship between 
integrity constraints and OO data modelling is 
discussed in general. Section 3 briefly reviews 
geographical modelling and object-oriented modelling. 
Section 4 gives an object-oriented data model for 
describing the geometrical structure of geographical 
entities. Section 5 is a proposition for geographical data 
modelling based on the model of section 4, along with 
a study of a possible integrity constraint in geographical 
application. The ideas of this paper are summarized 
and the future research is outlined in section 6. 
1.3 Some definitions 
The use of concepts such as geographical information 
and geographical database, spatial database, 
geographical and spatial objects, etc. is not always very 
consistent in the litterature. The meaning of some of 
these concepts is explained next, the way they are used 
in this paper. 
Geographical database is a data storage, in which data 
about real world entities is stored. The two concepts, 
457 
'geographical database' and ‘spatial database' are used 
interchangeably in this paper. Geographical entity is a 
real world entity, which is represented by a 
geographical object class in the database. The notion of 
geographical modelling is understood here as a process 
of describing geographical entities by describing their 
geometrical and thematic properties. 
2 INTEGRITY CONSTRAINTS OF A DATA MODEL 
By database semantic integrity is meant the techniques 
used to keep the database in a consistent state with 
respect to the constraints specified on the database 
(Elmasri & Navathe, 1989, p.589). In order to prevent 
an inconsistent database state, whenever an update is 
applied to the database, the semantic integrity checking 
part of the system determines if the update will cause a 
constraint violation. The term integrity is used to 
cover semantic integrity and transaction integrity. 
Semantic integrity emphasizes the meaning of the 
integrity constraint, while transction integrity is related 
to concurrency control and database recovery 
techniques. This paper focuses on the semantic 
interpretation of the concept. 
Integrity constraints constitute a set of rules that say 
what the allowable states of individual objects in the 
database are, and thus what the allowable states of the 
whole database are. In a database system, the 
specification of data types is presented in the database 
schema as a description of the data content in the 
database. This specification of data in itself constrains 
the database: when an update operation is performed 
in the database, be it insert, delete or change of some 
instance of a data type, the instance is constrained to be 
correct according to the data definition. This is implicit 
integrity constraint of a data model, i.e. constraint that 
is included in the schema. For example, when in a 
relational schema some relation attribute is specified 
as a key attribute, a check is made during database 
update operations that no two tuples having the same 
key attribute value are inserted into the database. Other 
example of'implicit relational model constraints is 
referential integrity, i.e. tuple value that is not stored 
in the database cannot be referred to. Unfortunately, 
many of the relational model implementations do not 
enforce this constraint according to (Elmasri & 
Navathe, 1989, p.145). 
What is not implicit in the data model has to be 
specified explicitly, outside the database schema. In 
other words, what you can't specify using the schema 
constructs, you have to program yourself. For example, 
in the information system built up above a relational 
database the before mentioned referential integrity 
constraints have to be enforced in the application 
program. The semantics of the data is thus embedded 
in code that is not part of the database, meaning that it 
is not managed, that is protected from inconsistencies, 
as other data on a real-world entity. The problem is 
that several different integrity constraints might be 
specified to the database, and it is up to the application 
programmer to be sure that no two conflicting 
constraints are specified.